Status: [LIVE] Endpoints found: https://www.pexels.com/api/ Source: https://github.com/GitHub-Zho/china_video_bot # 🇨🇳 China Video Bot An automated multi-agent pipeline that generates and publishes short-form China travel & culture videos to YouTube and Instagram Reels — daily, with zero manual intervention. ## What it does Every day at 9:00 AM UTC the bot: 1. **Generates a script** via Groq (Llama 3.3 70B, free tier) — 60-90 second voiceover copy targeting English-speaking audiences curious about China 2. **Downloads beautiful images** from Pexels + Unsplash (free APIs) — China landscapes, cities, food, culture 3. **Creates a voiceover + subtitles** via Microsoft Edge TTS (free) — natural English narration with precisely-timed SRT subtitles 4. **Assembles the video** via MoviePy + FFmpeg — 1920×1080 YouTube version and 1080×1920 Instagram Reels version, with burned-in subtitles 5. **Uploads to YouTube** via YouTube Data API v3 with automatic title, description, and tags 6. **Collects analytics** 3 days after publish and feeds performance data back to improve future scripts ## Two audience types | Audience | Description | |---|---| | **Explorer** | People who've seen China content and want hidden gems, lesser-known destinations, surprising facts | | **Newcomer** | People curious about China but don't know where to start — first-timer perspective | ## Architecture ``` Orchestrator ├── Script Agent — Groq API (Llama 3.3 70B) + template fallback ├── Image Agent — Pexels API + Unsplash API (alternating) ├── Voice Agent — edge-tts (Microsoft, free) → MP3 + SRT ├── Video Agent — MoviePy slideshow + FFmpeg subtitle burn ├── Publisher Agent — YouTube Data API v3 (OAuth2, refresh token) └── Analytics Agent — YouTube Analytics API v2 → feedback loop ``` ## Tech stack — 100% free APIs | Component | Tool | Cost | |---|---|---| | LLM / Script | [Groq](https://console.groq.com) (Llama 3.3 70B) | Free | | Images | Pexels API + Unsplash API | Free | | Text-to-speech | Microsoft Edge TTS (`edge-tt